Artificial Intelligence for Diabetic Retinopathy Screening Using Color Retinal Photographs: From Development to Deployment.

Abstract in brief

• This article reviews how artificial intelligence (AI), especially deep learning, is being developed and used to detect diabetic retinopathy (DR) from retinal photographs, aiming to reduce the manual effort and resources needed for screening.
• Early AI methods used machine learning with feature extraction, showing high sensitivity but lower specificity, while deep learning improved both measures, leading to FDA approval for autonomous DR screening, although real-world implementation is still limited.
• Challenges for widespread AI deployment include technical integration with health record systems, workflow adjustments, ethical concerns like data privacy, and acceptance by healthcare providers and patients, with some potential benefits like increased screening and referrals yet to be conclusively proven.